library(dplyr)
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(skimr)
library(corrplot)
## corrplot 0.84 loaded
library(psych)
library(ggplot2)
## 
## Attaching package: 'ggplot2'
## The following objects are masked from 'package:psych':
## 
##     %+%, alpha
library(patchwork)
library(reshape2)

source("cor_regress.R")
load('data/load_effects_DFR.RData')
load('data/behav.RData')
load('data/connectivity_data.RData')
load('data/structural_measures.RData')


scanner_cov <- p200_demographics[,c(1,4)]

col <- colorRampPalette(c("dark blue","light blue", "light green", "yellow", "orange", "red"))(200)

Behavioral Correlations

First, let’s look at the relationship between accuracy on the DFR task and our constructs. People with higher capacity do better. This is a nice sanity check. We’re not going to include the scanner as the covariate because there’s no need to think there’s an issue because none of the measures are fMRI measures.

# first, make sure we've got only subjects that have fMRI data 
DFR_acc <- p200_data[p200_data$PTID %in% p200_indiv_ROI_DFR_delay$PTID,c(1,7:8)]
DFR_acc$LE <- DFR_acc$XDFR_MRI_ACC_L3 - DFR_acc$XDFR_MRI_ACC_L1
constructs_fMRI <- construct_vars_omnibus[construct_vars_omnibus$PTID %in% p200_indiv_ROI_DFR_delay$PTID,]

acc_constructs_corr <- corr.test(DFR_acc[,2:4],constructs_fMRI[,2:7])

corrplot(acc_constructs_corr$r, addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(DFR_acc, constructs_fMRI)

ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = XDFR_MRI_ACC_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("Relationship between DFR L3 accuracy and omnibus WM span")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$omnibus_span_no_DFR_MRI)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$omnibus_span_no_DFR_MRI
## t = 3.9312, df = 168, p-value = 0.0001234
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.1461113 0.4223081
## sample estimates:
##       cor 
## 0.2902426
ggplot(data_for_plot, aes(y=XDFR_MRI_ACC_L1, x = omnibus_span_no_DFR_MRI))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("Relationship between DFR L1 accuracy and omnibus WM span")

cor.test(data_for_plot$XDFR_MRI_ACC_L1,data_for_plot$omnibus_span_no_DFR_MRI)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L1 and data_for_plot$omnibus_span_no_DFR_MRI
## t = 3.1219, df = 168, p-value = 0.002116
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.08670135 0.37157808
## sample estimates:
##       cor 
## 0.2341599

Now, let’s look overall at the the relationships between our constructs and the clinical variables.

constructs_clin_corr <- corr.test(construct_vars_omnibus[,2:7], p200_clinical_zscores[,2:13])

rownames(constructs_clin_corr$r) <- c("Simple Span", "Complex Span", "GM/IC", "LTM", "Intelligence", "Omnibus Span")
corrplot(constructs_clin_corr$r, addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

We’re mostly theoretically interested in the relationship between omnibus span (our working memory capacity measure) and measures like WHODAS and BPRS, which get at general psychiatric dysfunction. Let’s plot those specifically, just to look at the relationships at a little moe closely.

Both variables show a negative relationship, with lower levels of psychiatric distress at higher WM capacity. As we saw above, the relationship with BPRS Total is statistically significant at p < 0.01, but WHODAS is not.

data_for_plot <- merge(construct_vars_omnibus,p200_clinical_zscores)

ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = WHO_ST_S32))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("Relationship between omnibus span and WHODAS")

ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = BPRS_TOT))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("Relationship between omnibus span and BPRS -- Total")

Behavioral to Load Effects

DFR ROIs

Now, we want to look at relationships between behavior and load effects. Not much to see here.

constructs_load_effects <- cor_regress(construct_vars_omnibus[,1:7],p200_indiv_ROI_DFR_delay,scanner_cov)

rownames(constructs_load_effects[["results"]]) <- c("Simple Span", "Complex Span", "GM/IC", "LTM", "Intelligence", "Omnibus Span")
corrplot(as.matrix(constructs_load_effects[["results"]]), addCoef.col = "black", number.cex=0.5, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

For good measure, let’s just look at some scatter plots. We’re looking at 3 examples here - all the ROIs together, the L dlPFC and L IPS, regions that show particularly strong load effects and that also have theoretical implications with working memory. No wonder we didn’t see any significant correlations - it looks like there’s a non-linear relationship here. We should explore this more.

data_for_plot <- merge(p200_indiv_ROI_DFR_delay,constructs_fMRI)

p1 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = DFR_ROIs))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = DFR_L_dlPFC))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = DFR_L_IPS))+
  geom_point()+
  stat_smooth(method="loess")


p4 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = DFR_L_preSMA))+
  geom_point()+
  stat_smooth(method="loess")

(p1+p2)/(p3+p4)

data_for_plot <- merge(data_for_plot,p200_demographics)
model <- lm(DFR_L_dlPFC ~ omnibus_span_no_DFR_MRI + SCANNER, data = data_for_plot)

x = data.frame(resid = residuals(model), fitted = fitted(model))

ggplot(data=x,aes(x=resid))+
  geom_histogram(bins = 30)

ggplot(data = x, aes(x=fitted,y=resid))+
  geom_point()+
  stat_smooth(method="loess")

Now let’s look at these same relationships with accuracy. Interestingly, these show a significant linear relationship, which is different than with capacity.

cog_fMRI <- p200_cog_zscores[p200_cog_zscores$PTID %in% p200_indiv_ROI_DFR_delay$PTID,]

data_for_plot <- merge(cog_fMRI,p200_indiv_ROI_DFR_delay)

p1 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = DFR_ROIs))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=1,y=2.5,label="r=0.24**")


cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$DFR_ROIs)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$DFR_ROIs
## t = 3.3308, df = 168, p-value = 0.001065
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.1022032 0.3849811
## sample estimates:
##       cor 
## 0.2488888
p2 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = DFR_L_dlPFC))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=1,y=1,label="r=0.26***")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$DFR_L_dlPFC)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$DFR_L_dlPFC
## t = 3.5387, df = 168, p-value = 0.0005202
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.1175198 0.3981079
## sample estimates:
##       cor 
## 0.2633753
p3 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = DFR_L_IPS))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=1,y=1,label="r=0.24**")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$DFR_L_IPS)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$DFR_L_IPS
## t = 3.2651, df = 168, p-value = 0.001326
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.09734466 0.38079319
## sample estimates:
##       cor 
## 0.2442798
p1 + p2 + p3

HPC

Only thing signficant linear relationship is accuracy @ L3 with R posterior HPC delay @ L3.

data_for_plot <- merge(cog_fMRI,p200_HPC_Ant)

p1 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Anterior Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$R_DELAY_L3
## t = -0.24579, df = 167, p-value = 0.8061
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.1694896  0.1323238
## sample estimates:
##         cor 
## -0.01901613
p2 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Anterior Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$L_DELAY_L3
## t = -0.66747, df = 167, p-value = 0.5054
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.2009771  0.1001579
## sample estimates:
##       cor 
## -0.051582
data_for_plot <- merge(cog_fMRI,p200_HPC_Med)

p3 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Medial Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$R_DELAY_L3
## t = -0.85835, df = 167, p-value = 0.3919
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.21508328  0.08554072
## sample estimates:
##         cor 
## -0.06627514
p4 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Medial Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$L_DELAY_L3
## t = -0.7182, df = 167, p-value = 0.4736
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.20473555  0.09627596
## sample estimates:
##         cor 
## -0.05549056
data_for_plot <- merge(cog_fMRI,p200_HPC_Post)

p5 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Posterior Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$R_DELAY_L3
## t = -1.9953, df = 167, p-value = 0.04763
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.296718947 -0.001672515
## sample estimates:
##       cor 
## -0.152594
p6 <- ggplot(data_for_plot, aes(x=XDFR_MRI_ACC_L3, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Posterior Delay L3")

cor.test(data_for_plot$XDFR_MRI_ACC_L3,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$XDFR_MRI_ACC_L3 and data_for_plot$L_DELAY_L3
## t = -1.4181, df = 167, p-value = 0.158
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.25582732  0.04258127
## sample estimates:
##        cor 
## -0.1090799
(p1 + p3 + p5)/(p2 + p4 + p6)

No significant relatinonship between HPC activity and omnibus span.

data_for_plot <- merge(constructs_fMRI,p200_HPC_Ant)

p1 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Anterior Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$R_DELAY_L3
## t = 0.90723, df = 167, p-value = 0.3656
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.08179362  0.21867930
## sample estimates:
##        cor 
## 0.07003112
p2 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Anterior Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$L_DELAY_L3
## t = 0.25867, df = 167, p-value = 0.7962
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.1313440  0.1704579
## sample estimates:
##        cor 
## 0.02001285
data_for_plot <- merge(constructs_fMRI,p200_HPC_Med)

p3 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Medial Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$R_DELAY_L3
## t = -0.18199, df = 167, p-value = 0.8558
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.1646912  0.1371704
## sample estimates:
##         cor 
## -0.01408122
p4 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Medial Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$L_DELAY_L3
## t = 0.017239, df = 167, p-value = 0.9863
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.1496562  0.1522634
## sample estimates:
##        cor 
## 0.00133401
data_for_plot <- merge(constructs_fMRI,p200_HPC_Post)

p5 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = R_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("R HPC Posterior Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$R_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$R_DELAY_L3
## t = -0.49673, df = 167, p-value = 0.62
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.1882783  0.1132064
## sample estimates:
##         cor 
## -0.03841002
p6 <- ggplot(data_for_plot, aes(x=omnibus_span_no_DFR_MRI, y = L_DELAY_L3))+
  geom_point()+
  stat_smooth(method="lm")+
  ggtitle("L HPC Posterior Delay L3")

cor.test(data_for_plot$omnibus_span_no_DFR_MRI,data_for_plot$L_DELAY_L3)
## 
##  Pearson's product-moment correlation
## 
## data:  data_for_plot$omnibus_span_no_DFR_MRI and data_for_plot$L_DELAY_L3
## t = -0.97798, df = 167, p-value = 0.3295
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.22387256  0.07636713
## sample estimates:
##         cor 
## -0.07546286
(p1 + p3 + p5)/(p2 + p4 + p6)

Nothing significant with clinical/load effects.

clinical_load_corr <- cor_regress(p200_clinical_zscores,p200_indiv_ROI_DFR_delay,scanner_cov)

corrplot(as.matrix(clinical_load_corr[["results"]]),addCoef.col = "black", number.cex=0.45, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

But, let’s just look at some scatter plots, just in case it’s a non-linear relationship like with capacity. Specifically, let’s look at BPRS because it’s the measure of global psychiatric dysfuntion that showed a relationship with capacity, and WHODAS, because it’s another general measure. Neither of these show any relationship.

data_for_plot <- merge(p200_indiv_ROI_DFR_delay,p200_clinical_zscores)

p1 <- ggplot(data_for_plot, aes(x=BPRS_TOT, y = DFR_ROIs))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot, aes(x=BPRS_TOT, y = DFR_L_dlPFC))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot, aes(x=BPRS_TOT, y = DFR_L_IPS))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot, aes(x=WHO_ST_S32, y = DFR_ROIs))+
  geom_point()+
  stat_smooth(method="loess")

p5<- ggplot(data_for_plot, aes(x=WHO_ST_S32, y = DFR_L_dlPFC))+
  geom_point()+
  stat_smooth(method="loess")

p6 <- ggplot(data_for_plot, aes(x=WHO_ST_S32, y = DFR_L_IPS))+
  geom_point()+
  stat_smooth(method="loess")

(p1 + p2 + p3)/(p4 + p5 + p6)

Resting State Connectivity

Within/Across Regions

This data is averaged over networks, and included global signal regression and detrending.

RS_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7],p200_RS, scanner_cov)
corrplot(as.matrix(RS_to_constructs_corr[["results"]]), addCoef.col = "black", number.cex=0.45, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

Let’s just investigate some of the more interesting ones - FPCN is theoretically related to WM, so let’s look at those scatter plots to make sure there isn’t a relationship like there is with load effects.

data_for_plot <- merge(construct_vars_omnibus,p200_RS)
p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`FrontoParietal/Default`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.4,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,5],digits=2)))

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`FrontoParietal/Visual`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.25,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,10],digits=2)))

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`FrontoParietal/CinguloOperc`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.4,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,14],digits=2)))

p4 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`FrontoParietal/DorsalAttn`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.4,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,17],digits=2)))

p5 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`FrontoParietal/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.23,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,19],digits=2)))

p6 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=`VentralAttn/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=0.5,y=0.35,label=paste("r =",round(RS_to_constructs_corr[["results"]][6,20],digits=2)))

(p1+p2+p3)/(p4+p5+p6)+
  plot_annotation(title = "FPCN connectivity with omnibus span")

RS_to_clin_corr <- cor_regress(p200_clinical_zscores,p200_RS, scanner_cov)
corrplot(as.matrix(RS_to_clin_corr[["results"]]), addCoef.col = "black", number.cex=0.3,method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_clinical_zscores,p200_RS)
p1 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`FrontoParietal/Default`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.3,label="r=-0.13")

p2 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`FrontoParietal/Visual`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.2,label="r=0.06")

p3 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`FrontoParietal/CinguloOperc`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.4,label="r=0.13")

p4 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`FrontoParietal/DorsalAttn`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.35,label="r=0.02")

p5 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`FrontoParietal/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.225,label="r=-0.06")

p6 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=`VentralAttn/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=3,y=.35,label="r=0.04")

(p1+p2+p3)/(p4+p5+p6)+
  plot_annotation(title="FPCN/BPRS")

p1 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`FrontoParietal/Default`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=2,y=.4,label="r=-0.02")

p2 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`FrontoParietal/Visual`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=1.75,y=.25,label="r=0.07")

p3 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`FrontoParietal/CinguloOperc`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=2,y=.4,label="r=0.10")

p4 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`FrontoParietal/DorsalAttn`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=1.5,y=.35,label="r=-0.01")

p5 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`FrontoParietal/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=2,y=.225,label="r=-0.07")

p6 <- ggplot(data_for_plot,aes(x=WHO_ST_S32,y=`VentralAttn/FrontoParietal`))+
  geom_point()+
  stat_smooth(method="lm")+
  geom_text(x=2,y=.3,label="r=0.05")

(p1+p2+p3)/(p4+p5+p6)+
  plot_annotation(title="FPCN/WHODAS")

BCT Measures

BCT_to_constructs_corr <- cor_regress(p200_BCT_forCorr,construct_vars_omnibus[,1:7], scanner_cov)
corrplot(as.matrix(BCT_to_constructs_corr[["results"]]), addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_BCT_forCorr,construct_vars_omnibus)

p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Participation_Coef_Mean))+
  geom_point()+
  stat_smooth(method="lm")

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Global_Eff))+
  geom_point()+
  stat_smooth(method="lm")

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Modularity_Louvain_N))+
  geom_point()+
  stat_smooth(method="lm")

p1+p2+p3+
  plot_annotation(title = "Omnibus Span/BCT measures")

BCT_to_clinical_corr <- cor_regress(p200_BCT_forCorr,p200_clinical_zscores, scanner_cov)
corrplot(as.matrix(BCT_to_clinical_corr[["results"]]),addCoef.col = "black", number.cex=0.5, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

Individual Network Participation Coefficients

indiv_partic_coeff_to_constructs_corr <- cor_regress(p200_indiv_network_ParticCoeff,construct_vars_omnibus[,1:7], scanner_cov)
corrplot(as.matrix(indiv_partic_coeff_to_constructs_corr[["results"]]), addCoef.col = "black", number.cex=0.75,method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_indiv_network_ParticCoeff,construct_vars_omnibus)

ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FrontoParietal))+
  geom_point()+
  stat_smooth(method="lm")

indiv_partic_coeff_to_clin_corr <- cor_regress(p200_indiv_network_ParticCoeff,p200_clinical_zscores, scanner_cov)
corrplot(as.matrix(indiv_partic_coeff_to_clin_corr[["results"]]), addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_indiv_network_ParticCoeff,p200_clinical_zscores)

ggplot(data=data_for_plot,aes(x=BPRS_TOT,y=FrontoParietal))+
  geom_point()+
  stat_smooth(method="lm")

ggplot(data=data_for_plot,aes(x=WHO_ST_S32,y=FrontoParietal))+
  geom_point()+
  stat_smooth(method="lm")
## Warning: Removed 1 rows containing non-finite values (stat_smooth).
## Warning: Removed 1 rows containing missing values (geom_point).

Beta Series Connectivity

This analysis uses the entire p200 dataset (at least those that have useable fMRI data) but uses ROIs derived from the HCP n-back task instead of from our own data.

Cue

BS_cue_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], p200_beta_conn_cue,scanner_cov)
corrplot(as.matrix(BS_cue_to_constructs_corr[["results"]]),addCoef.col = "black", number.cex=0.5, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(construct_vars_omnibus,p200_beta_conn_cue)

p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FPCN_L3))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_HPC_L3))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=HPC_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity at high load during cue with Omnibus Span")

p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FPCN_LE))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_HPC_LE))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=HPC_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity Load Effects during cue with Omnibus Span")

BS_cue_to_clinical_corr <- cor_regress(p200_clinical_zscores, p200_beta_conn_cue,scanner_cov)
corrplot(as.matrix(BS_cue_to_clinical_corr[["results"]]), addCoef.col = "black", number.cex=0.5,method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_clinical_zscores,p200_beta_conn_cue)

p1 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FPCN_L3))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_HPC_L3))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=HPC_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity at high load during cue with BPRS")

p1 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FPCN_LE))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_HPC_LE))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=HPC_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity Load Effects during cue with BPRS")

Delay

BS_delay_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], p200_beta_conn_delay,scanner_cov)
corrplot(as.matrix(BS_delay_to_constructs_corr[["results"]]), addCoef.col = "black", number.cex=0.5, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(construct_vars_omnibus,p200_beta_conn_delay)

p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FPCN_L3))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_HPC_L3))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=HPC_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity at high load during Delay with Omnibus Span")

p1 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FPCN_LE))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_HPC_LE))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=FPCN_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=HPC_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity Load Effects during Delay with Omnibus Span")

BS_delay_to_clinical_corr <- cor_regress(p200_clinical_zscores, p200_beta_conn_delay,scanner_cov)
corrplot(as.matrix(BS_delay_to_clinical_corr[["results"]]), addCoef.col = "black", number.cex=0.5, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(p200_clinical_zscores,p200_beta_conn_delay)

p1 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FPCN_L3))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_HPC_L3))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=HPC_FFA_L3))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity at high load during delay with BPRS")

p1 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FPCN_LE))+
  geom_point()+
  stat_smooth(method="loess")

p2 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_HPC_LE))+
  geom_point()+
  stat_smooth(method="loess")

p3 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=FPCN_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

p4 <- ggplot(data_for_plot,aes(x=BPRS_TOT,y=HPC_FFA_LE))+
  geom_point()+
  stat_smooth(method="loess")

(p1 +p2)/(p3+p4)+
  plot_annotation(title="Beta Series Connectivity Load Effects during delay with BPRS")

Structural Connectivity

FA_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], FA_Data,scanner_cov)
corrplot(as.matrix(FA_to_constructs_corr[["results"]]), method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

FA_to_clin_corr <- cor_regress(FA_Data,p200_clinical_zscores, scanner_cov)
corrplot(as.matrix(FA_to_clin_corr[["results"]]),method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

Cortical Thickness

cortical_DFR_thick_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], p200_DFR_fullMask_cortical_thickness,scanner_cov)
corrplot(as.matrix(cortical_DFR_thick_to_constructs_corr[["results"]]),addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

data_for_plot <- merge(constructs_fMRI,p200_DFR_fullMask_cortical_thickness)

cue_RH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Cue_RH))+
  geom_point()+
  stat_smooth(method="loess")

cue_LH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Cue_LH))+
  geom_point()+
  stat_smooth(method="loess")

delay_RH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Delay_RH))+
  geom_point()+
  stat_smooth(method="loess")

delay_LH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Delay_LH))+
  geom_point()+
  stat_smooth(method="loess")

probe_RH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Probe_RH))+
  geom_point()+
  stat_smooth(method="loess")

probe_LH <- ggplot(data=data_for_plot,aes(x=omnibus_span_no_DFR_MRI,y=Probe_LH))+
  geom_point()+
  stat_smooth(method="loess")


(cue_RH + delay_RH + probe_RH) / (cue_LH + delay_LH + probe_LH)+ 
  plot_annotation(title="cortical thickness from DFR delay period clusters vs omnibus span")
## Warning: Removed 1 rows containing non-finite values (stat_smooth).
## Warning: Removed 1 rows containing missing values (geom_point).

cortical_DFR_thick_to_clinical_corr <- cor_regress(p200_DFR_fullMask_cortical_thickness,p200_clinical_zscores, scanner_cov)
corrplot(as.matrix(cortical_DFR_thick_to_clinical_corr[["results"]]),addCoef.col = "black", number.cex=0.75, method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

cortical_thick_LH_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], aparc_LH_MTHICK,scanner_cov)
corrplot(as.matrix(cortical_thick_LH_to_constructs_corr[["results"]]), method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

cortical_thick_LH_to_clinical_corr <- cor_regress(p200_clinical_zscores, aparc_LH_MTHICK,scanner_cov)
corrplot(as.matrix(cortical_thick_LH_to_clinical_corr[["results"]]), method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

cortical_thick_RH_to_constructs_corr <- cor_regress(construct_vars_omnibus[,1:7], aparc_RH_MTHICK,scanner_cov)
corrplot(as.matrix(cortical_thick_RH_to_constructs_corr[["results"]]), method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)

cortical_thick_RH_to_clinical_corr <- cor_regress(p200_clinical_zscores, aparc_RH_MTHICK,scanner_cov)
corrplot(as.matrix(cortical_thick_RH_to_clinical_corr[["results"]]), method="color", tl.col = "black", tl.cex = 1, col = col, sig.level = 0.05)